Metanalysis of performance on women caused by (stFemale)

Geiser C. Challco geiser@alumni.usp.br

Initial Variables and Loading Data

env <- "stFemale"
gender <- "women"
to_remove <- c('S11')
sub.groups <- c("country","age","ed.level","intervention",
                "country:age","country:ed.level","country:intervention",
                "age:intervention","ed.level:intervention",
                "country:age:intervention","country:ed.level:intervention")
dat <- read_excel("../data/data-without-outliers.xlsx", sheet = "perform-env.gender-descriptive")
dat <- dat[!dat$study %in% to_remove, ]

leg <- read_excel("../data/data-without-outliers.xlsx", sheet = "legend")
## New names:
## • `` -> `...10`
leg <- leg[!leg$study %in% to_remove, ]

idx.e <- which(dat$env == env & dat$gender == gender)
idx.c <- which(dat$env == "control" & dat$gender == gender)

data <- data.frame(
  study = dat$study[idx.c],
  n.e = dat$N[idx.e], mean.e = dat$M[idx.e], sd.e = dat$SD[idx.e],
  n.c = dat$N[idx.c], mean.c = dat$M[idx.c], sd.c = dat$SD[idx.c]
)
for (cgroups in strsplit(sub.groups,":")) {
  data[[paste0(cgroups, collapse = ":")]] <- sapply(data$study, FUN = function(x) {
    paste0(sapply(cgroups, FUN = function(namecol) leg[[namecol]][which(x == leg$study)]), collapse = ":")
  })
}
data[["lbl"]] <- sapply(data$study, FUN = function(x) leg$Note[which(x == leg$study)])

Perform meta-analyses

m.cont <- metacont(
  n.e = n.e, mean.e = mean.e, sd.e = sd.e, n.c = n.c, mean.c = mean.c, sd.c = sd.c,
  studlab = lbl, data = data, sm = "SMD", method.smd = "Hedges",
  fixed = F, random = T, method.tau = "REML", hakn = T, title = paste("Performance for",gender,"in",env)
)
summary(m.cont)
## Review:     Performance for women in stFemale
## 
##                                    SMD            95%-CI %W(random)
## S1                              0.0337 [-0.5959; 0.6633]       10.9
## S2                             -0.0214 [-0.5446; 0.5018]       14.8
## S3                             -0.5787 [-1.3128; 0.1555]        8.3
## S4                              0.6745 [-0.1048; 1.4539]        7.4
## S5                              0.2265 [-0.3774; 0.8305]       11.7
## S6                              0.4025 [-0.3214; 1.1264]        8.5
## S7                              0.5444 [-0.0754; 1.1641]       11.2
## S8: Conducted by BNU           -0.1893 [-0.8121; 0.4335]       11.1
## S9: Albuquerque, et al. (2017) -0.4198 [-1.1460; 0.3064]        8.5
## S10: Only use prompt msgs       0.1888 [-0.5722; 0.9498]        7.8
## 
## Number of studies combined: k = 10
## Number of observations: o = 366
## 
##                         SMD            95%-CI    t p-value
## Random effects model 0.0823 [-0.1890; 0.3537] 0.69  0.5098
## 
## Quantifying heterogeneity:
##  tau^2 = 0.0169 [0.0000; 0.4121]; tau = 0.1299 [0.0000; 0.6420]
##  I^2 = 20.0% [0.0%; 60.5%]; H = 1.12 [1.00; 1.59]
## 
## Test of heterogeneity:
##      Q d.f. p-value
##  11.25    9  0.2587
## 
## Details on meta-analytical method:
## - Inverse variance method
## - Restricted maximum-likelihood estimator for tau^2
## - Q-profile method for confidence interval of tau^2 and tau
## - Hartung-Knapp adjustment for random effects model
## - Hedges' g (bias corrected standardised mean difference; using exact formulae)
forest(m.cont, digits=2, digits.sd = 2, test.overall = T, label.e = paste0(gender,':',env))

Subgroup analysis by “country”

m.sg4sub <- update.meta(m.cont, subgroup = country, random = T, fixed = F)
summary(m.sg4sub)
## Review:     Performance for women in stFemale
## 
##                                    SMD            95%-CI %W(random) country
## S1                              0.0337 [-0.5959; 0.6633]       10.9  Brazil
## S2                             -0.0214 [-0.5446; 0.5018]       14.8  Brazil
## S3                             -0.5787 [-1.3128; 0.1555]        8.3  Brazil
## S4                              0.6745 [-0.1048; 1.4539]        7.4  Brazil
## S5                              0.2265 [-0.3774; 0.8305]       11.7  Brazil
## S6                              0.4025 [-0.3214; 1.1264]        8.5  Brazil
## S7                              0.5444 [-0.0754; 1.1641]       11.2  Brazil
## S8: Conducted by BNU           -0.1893 [-0.8121; 0.4335]       11.1   China
## S9: Albuquerque, et al. (2017) -0.4198 [-1.1460; 0.3064]        8.5  Brazil
## S10: Only use prompt msgs       0.1888 [-0.5722; 0.9498]        7.8  Brazil
## 
## Number of studies combined: k = 10
## Number of observations: o = 366
## 
##                         SMD            95%-CI    t p-value
## Random effects model 0.0823 [-0.1890; 0.3537] 0.69  0.5098
## 
## Quantifying heterogeneity:
##  tau^2 = 0.0169 [0.0000; 0.4121]; tau = 0.1299 [0.0000; 0.6420]
##  I^2 = 20.0% [0.0%; 60.5%]; H = 1.12 [1.00; 1.59]
## 
## Test of heterogeneity:
##      Q d.f. p-value
##  11.25    9  0.2587
## 
## Results for subgroups (random effects model):
##                    k     SMD            95%-CI  tau^2    tau     Q   I^2
## country = Brazil   9  0.1162 [-0.1849; 0.4173] 0.0238 0.1542 10.44 23.3%
## country = China    1 -0.1893 [-0.8121; 0.4335]     --     --  0.00    --
## 
## Test for subgroup differences (random effects model):
##                     Q d.f. p-value
## Between groups   0.79    1  0.3738
## 
## Details on meta-analytical method:
## - Inverse variance method
## - Restricted maximum-likelihood estimator for tau^2
## - Q-profile method for confidence interval of tau^2 and tau
## - Hartung-Knapp adjustment for random effects model
## - Hedges' g (bias corrected standardised mean difference; using exact formulae)
forest(m.sg4sub, digits=2, digits.sd = 2, test.overall = T, label.e = paste0(gender,':',env))

Subgroup analysis by “age”

m.sg4sub <- update.meta(m.cont, subgroup = age, random = T, fixed = F)
summary(m.sg4sub)
## Review:     Performance for women in stFemale
## 
##                                    SMD            95%-CI %W(random)            age
## S1                              0.0337 [-0.5959; 0.6633]       10.9     adolescent
## S2                             -0.0214 [-0.5446; 0.5018]       14.8     adolescent
## S3                             -0.5787 [-1.3128; 0.1555]        8.3     adolescent
## S4                              0.6745 [-0.1048; 1.4539]        7.4          adult
## S5                              0.2265 [-0.3774; 0.8305]       11.7          adult
## S6                              0.4025 [-0.3214; 1.1264]        8.5          adult
## S7                              0.5444 [-0.0754; 1.1641]       11.2          adult
## S8: Conducted by BNU           -0.1893 [-0.8121; 0.4335]       11.1        unknown
## S9: Albuquerque, et al. (2017) -0.4198 [-1.1460; 0.3064]        8.5 no-restriction
## S10: Only use prompt msgs       0.1888 [-0.5722; 0.9498]        7.8     adolescent
## 
## Number of studies combined: k = 10
## Number of observations: o = 366
## 
##                         SMD            95%-CI    t p-value
## Random effects model 0.0823 [-0.1890; 0.3537] 0.69  0.5098
## 
## Quantifying heterogeneity:
##  tau^2 = 0.0169 [0.0000; 0.4121]; tau = 0.1299 [0.0000; 0.6420]
##  I^2 = 20.0% [0.0%; 60.5%]; H = 1.12 [1.00; 1.59]
## 
## Test of heterogeneity:
##      Q d.f. p-value
##  11.25    9  0.2587
## 
## Results for subgroups (random effects model):
##                        k     SMD            95%-CI tau^2 tau    Q  I^2
## age = adolescent       4 -0.0759 [-0.5432; 0.3914]     0   0 2.42 0.0%
## age = adult            4  0.4401 [ 0.1343; 0.7459]     0   0 0.95 0.0%
## age = unknown          1 -0.1893 [-0.8121; 0.4335]    --  -- 0.00   --
## age = no-restriction   1 -0.4198 [-1.1460; 0.3064]    --  -- 0.00   --
## 
## Test for subgroup differences (random effects model):
##                      Q d.f. p-value
## Between groups   13.94    3  0.0030
## 
## Details on meta-analytical method:
## - Inverse variance method
## - Restricted maximum-likelihood estimator for tau^2
## - Q-profile method for confidence interval of tau^2 and tau
## - Hartung-Knapp adjustment for random effects model
## - Hedges' g (bias corrected standardised mean difference; using exact formulae)
forest(m.sg4sub, digits=2, digits.sd = 2, test.overall = T, label.e = paste0(gender,':',env))

Subgroup analysis by “ed.level”

m.sg4sub <- update.meta(m.cont, subgroup = ed.level, random = T, fixed = F)
summary(m.sg4sub)
## Review:     Performance for women in stFemale
## 
##                                    SMD            95%-CI %W(random)         ed.level
## S1                              0.0337 [-0.5959; 0.6633]       10.9  upper-secundary
## S2                             -0.0214 [-0.5446; 0.5018]       14.8  upper-secundary
## S3                             -0.5787 [-1.3128; 0.1555]        8.3  upper-secundary
## S4                              0.6745 [-0.1048; 1.4539]        7.4 higher-education
## S5                              0.2265 [-0.3774; 0.8305]       11.7 higher-education
## S6                              0.4025 [-0.3214; 1.1264]        8.5 higher-education
## S7                              0.5444 [-0.0754; 1.1641]       11.2          unknown
## S8: Conducted by BNU           -0.1893 [-0.8121; 0.4335]       11.1          unknown
## S9: Albuquerque, et al. (2017) -0.4198 [-1.1460; 0.3064]        8.5          unknown
## S10: Only use prompt msgs       0.1888 [-0.5722; 0.9498]        7.8  upper-secundary
## 
## Number of studies combined: k = 10
## Number of observations: o = 366
## 
##                         SMD            95%-CI    t p-value
## Random effects model 0.0823 [-0.1890; 0.3537] 0.69  0.5098
## 
## Quantifying heterogeneity:
##  tau^2 = 0.0169 [0.0000; 0.4121]; tau = 0.1299 [0.0000; 0.6420]
##  I^2 = 20.0% [0.0%; 60.5%]; H = 1.12 [1.00; 1.59]
## 
## Test of heterogeneity:
##      Q d.f. p-value
##  11.25    9  0.2587
## 
## Results for subgroups (random effects model):
##                               k     SMD            95%-CI  tau^2    tau    Q   I^2
## ed.level = upper-secundary    4 -0.0759 [-0.5432; 0.3914]      0      0 2.42  0.0%
## ed.level = higher-education   3  0.3970 [-0.1540; 0.9481]      0      0 0.79  0.0%
## ed.level = unknown            3 -0.0029 [-1.2549; 1.2491] 0.1440 0.3795 4.59 56.5%
## 
## Test for subgroup differences (random effects model):
##                     Q d.f. p-value
## Between groups   6.30    2  0.0429
## 
## Details on meta-analytical method:
## - Inverse variance method
## - Restricted maximum-likelihood estimator for tau^2
## - Q-profile method for confidence interval of tau^2 and tau
## - Hartung-Knapp adjustment for random effects model
## - Hedges' g (bias corrected standardised mean difference; using exact formulae)
forest(m.sg4sub, digits=2, digits.sd = 2, test.overall = T, label.e = paste0(gender,':',env))

Subgroup analysis by “intervention”

m.sg4sub <- update.meta(m.cont, subgroup = intervention, random = T, fixed = F)
summary(m.sg4sub)
## Review:     Performance for women in stFemale
## 
##                                    SMD            95%-CI %W(random)
## S1                              0.0337 [-0.5959; 0.6633]       10.9
## S2                             -0.0214 [-0.5446; 0.5018]       14.8
## S3                             -0.5787 [-1.3128; 0.1555]        8.3
## S4                              0.6745 [-0.1048; 1.4539]        7.4
## S5                              0.2265 [-0.3774; 0.8305]       11.7
## S6                              0.4025 [-0.3214; 1.1264]        8.5
## S7                              0.5444 [-0.0754; 1.1641]       11.2
## S8: Conducted by BNU           -0.1893 [-0.8121; 0.4335]       11.1
## S9: Albuquerque, et al. (2017) -0.4198 [-1.1460; 0.3064]        8.5
## S10: Only use prompt msgs       0.1888 [-0.5722; 0.9498]        7.8
##                                                                        intervention
## S1                             Gender-stereotype color, ranking, badges, and avatar
## S2                             Gender-stereotype color, ranking, badges, and avatar
## S3                             Gender-stereotype color, ranking, badges, and avatar
## S4                             Gender-stereotype color, ranking, badges, and avatar
## S5                             Gender-stereotype color, ranking, badges, and avatar
## S6                             Gender-stereotype color, ranking, badges, and avatar
## S7                             Gender-stereotype color, ranking, badges, and avatar
## S8: Conducted by BNU           Gender-stereotype color, ranking, badges, and avatar
## S9: Albuquerque, et al. (2017) Gender-stereotype color, ranking, badges, and avatar
## S10: Only use prompt msgs           Gender-stereotyped motivational message prompts
## 
## Number of studies combined: k = 10
## Number of observations: o = 366
## 
##                         SMD            95%-CI    t p-value
## Random effects model 0.0823 [-0.1890; 0.3537] 0.69  0.5098
## 
## Quantifying heterogeneity:
##  tau^2 = 0.0169 [0.0000; 0.4121]; tau = 0.1299 [0.0000; 0.6420]
##  I^2 = 20.0% [0.0%; 60.5%]; H = 1.12 [1.00; 1.59]
## 
## Test of heterogeneity:
##      Q d.f. p-value
##  11.25    9  0.2587
## 
## Results for subgroups (random effects model):
##                                                      k    SMD            95%-CI  tau^2    tau     Q   I^2
## intervention = Gender-stereotype color, rankin ...   9 0.0735 [-0.2336; 0.3805] 0.0318 0.1782 11.17 28.4%
## intervention = Gender-stereotyped motivational ...   1 0.1888 [-0.5722; 0.9498]     --     --  0.00    --
## 
## Test for subgroup differences (random effects model):
##                     Q d.f. p-value
## Between groups   0.08    1  0.7788
## 
## Details on meta-analytical method:
## - Inverse variance method
## - Restricted maximum-likelihood estimator for tau^2
## - Q-profile method for confidence interval of tau^2 and tau
## - Hartung-Knapp adjustment for random effects model
## - Hedges' g (bias corrected standardised mean difference; using exact formulae)
forest(m.sg4sub, digits=2, digits.sd = 2, test.overall = T, label.e = paste0(gender,':',env))

Subgroup analysis by “country:age”

m.sg4sub <- update.meta(m.cont, subgroup = `country:age`, random = T, fixed = F)
summary(m.sg4sub)
## Review:     Performance for women in stFemale
## 
##                                    SMD            95%-CI %W(random)           country:age
## S1                              0.0337 [-0.5959; 0.6633]       10.9     Brazil:adolescent
## S2                             -0.0214 [-0.5446; 0.5018]       14.8     Brazil:adolescent
## S3                             -0.5787 [-1.3128; 0.1555]        8.3     Brazil:adolescent
## S4                              0.6745 [-0.1048; 1.4539]        7.4          Brazil:adult
## S5                              0.2265 [-0.3774; 0.8305]       11.7          Brazil:adult
## S6                              0.4025 [-0.3214; 1.1264]        8.5          Brazil:adult
## S7                              0.5444 [-0.0754; 1.1641]       11.2          Brazil:adult
## S8: Conducted by BNU           -0.1893 [-0.8121; 0.4335]       11.1         China:unknown
## S9: Albuquerque, et al. (2017) -0.4198 [-1.1460; 0.3064]        8.5 Brazil:no-restriction
## S10: Only use prompt msgs       0.1888 [-0.5722; 0.9498]        7.8     Brazil:adolescent
## 
## Number of studies combined: k = 10
## Number of observations: o = 366
## 
##                         SMD            95%-CI    t p-value
## Random effects model 0.0823 [-0.1890; 0.3537] 0.69  0.5098
## 
## Quantifying heterogeneity:
##  tau^2 = 0.0169 [0.0000; 0.4121]; tau = 0.1299 [0.0000; 0.6420]
##  I^2 = 20.0% [0.0%; 60.5%]; H = 1.12 [1.00; 1.59]
## 
## Test of heterogeneity:
##      Q d.f. p-value
##  11.25    9  0.2587
## 
## Results for subgroups (random effects model):
##                                       k     SMD            95%-CI tau^2 tau    Q  I^2
## country:age = Brazil:adolescent       4 -0.0759 [-0.5432; 0.3914]     0   0 2.42 0.0%
## country:age = Brazil:adult            4  0.4401 [ 0.1343; 0.7459]     0   0 0.95 0.0%
## country:age = China:unknown           1 -0.1893 [-0.8121; 0.4335]    --  -- 0.00   --
## country:age = Brazil:no-restriction   1 -0.4198 [-1.1460; 0.3064]    --  -- 0.00   --
## 
## Test for subgroup differences (random effects model):
##                      Q d.f. p-value
## Between groups   13.94    3  0.0030
## 
## Details on meta-analytical method:
## - Inverse variance method
## - Restricted maximum-likelihood estimator for tau^2
## - Q-profile method for confidence interval of tau^2 and tau
## - Hartung-Knapp adjustment for random effects model
## - Hedges' g (bias corrected standardised mean difference; using exact formulae)
forest(m.sg4sub, digits=2, digits.sd = 2, test.overall = T, label.e = paste0(gender,':',env))

Subgroup analysis by “country:ed.level”

m.sg4sub <- update.meta(m.cont, subgroup = `country:ed.level`, random = T, fixed = F)
summary(m.sg4sub)
## Review:     Performance for women in stFemale
## 
##                                    SMD            95%-CI %W(random)        country:ed.level
## S1                              0.0337 [-0.5959; 0.6633]       10.9  Brazil:upper-secundary
## S2                             -0.0214 [-0.5446; 0.5018]       14.8  Brazil:upper-secundary
## S3                             -0.5787 [-1.3128; 0.1555]        8.3  Brazil:upper-secundary
## S4                              0.6745 [-0.1048; 1.4539]        7.4 Brazil:higher-education
## S5                              0.2265 [-0.3774; 0.8305]       11.7 Brazil:higher-education
## S6                              0.4025 [-0.3214; 1.1264]        8.5 Brazil:higher-education
## S7                              0.5444 [-0.0754; 1.1641]       11.2          Brazil:unknown
## S8: Conducted by BNU           -0.1893 [-0.8121; 0.4335]       11.1           China:unknown
## S9: Albuquerque, et al. (2017) -0.4198 [-1.1460; 0.3064]        8.5          Brazil:unknown
## S10: Only use prompt msgs       0.1888 [-0.5722; 0.9498]        7.8  Brazil:upper-secundary
## 
## Number of studies combined: k = 10
## Number of observations: o = 366
## 
##                         SMD            95%-CI    t p-value
## Random effects model 0.0823 [-0.1890; 0.3537] 0.69  0.5098
## 
## Quantifying heterogeneity:
##  tau^2 = 0.0169 [0.0000; 0.4121]; tau = 0.1299 [0.0000; 0.6420]
##  I^2 = 20.0% [0.0%; 60.5%]; H = 1.12 [1.00; 1.59]
## 
## Test of heterogeneity:
##      Q d.f. p-value
##  11.25    9  0.2587
## 
## Results for subgroups (random effects model):
##                                              k     SMD            95%-CI  tau^2    tau    Q   I^2
## country:ed.level = Brazil:upper-secundary    4 -0.0759 [-0.5432; 0.3914]      0      0 2.42  0.0%
## country:ed.level = Brazil:higher-education   3  0.3970 [-0.1540; 0.9481]      0      0 0.79  0.0%
## country:ed.level = Brazil:unknown            2  0.0816 [-6.0390; 6.2023] 0.3462 0.5884 3.92 74.5%
## country:ed.level = China:unknown             1 -0.1893 [-0.8121; 0.4335]     --     -- 0.00    --
## 
## Test for subgroup differences (random effects model):
##                     Q d.f. p-value
## Between groups   7.24    3  0.0646
## 
## Details on meta-analytical method:
## - Inverse variance method
## - Restricted maximum-likelihood estimator for tau^2
## - Q-profile method for confidence interval of tau^2 and tau
## - Hartung-Knapp adjustment for random effects model
## - Hedges' g (bias corrected standardised mean difference; using exact formulae)
forest(m.sg4sub, digits=2, digits.sd = 2, test.overall = T, label.e = paste0(gender,':',env))

Subgroup analysis by “country:intervention”

m.sg4sub <- update.meta(m.cont, subgroup = `country:intervention`, random = T, fixed = F)
summary(m.sg4sub)
## Review:     Performance for women in stFemale
## 
##                                    SMD            95%-CI %W(random)
## S1                              0.0337 [-0.5959; 0.6633]       10.9
## S2                             -0.0214 [-0.5446; 0.5018]       14.8
## S3                             -0.5787 [-1.3128; 0.1555]        8.3
## S4                              0.6745 [-0.1048; 1.4539]        7.4
## S5                              0.2265 [-0.3774; 0.8305]       11.7
## S6                              0.4025 [-0.3214; 1.1264]        8.5
## S7                              0.5444 [-0.0754; 1.1641]       11.2
## S8: Conducted by BNU           -0.1893 [-0.8121; 0.4335]       11.1
## S9: Albuquerque, et al. (2017) -0.4198 [-1.1460; 0.3064]        8.5
## S10: Only use prompt msgs       0.1888 [-0.5722; 0.9498]        7.8
##                                                                       country:intervention
## S1                             Brazil:Gender-stereotype color, ranking, badges, and avatar
## S2                             Brazil:Gender-stereotype color, ranking, badges, and avatar
## S3                             Brazil:Gender-stereotype color, ranking, badges, and avatar
## S4                             Brazil:Gender-stereotype color, ranking, badges, and avatar
## S5                             Brazil:Gender-stereotype color, ranking, badges, and avatar
## S6                             Brazil:Gender-stereotype color, ranking, badges, and avatar
## S7                             Brazil:Gender-stereotype color, ranking, badges, and avatar
## S8: Conducted by BNU            China:Gender-stereotype color, ranking, badges, and avatar
## S9: Albuquerque, et al. (2017) Brazil:Gender-stereotype color, ranking, badges, and avatar
## S10: Only use prompt msgs           Brazil:Gender-stereotyped motivational message prompts
## 
## Number of studies combined: k = 10
## Number of observations: o = 366
## 
##                         SMD            95%-CI    t p-value
## Random effects model 0.0823 [-0.1890; 0.3537] 0.69  0.5098
## 
## Quantifying heterogeneity:
##  tau^2 = 0.0169 [0.0000; 0.4121]; tau = 0.1299 [0.0000; 0.6420]
##  I^2 = 20.0% [0.0%; 60.5%]; H = 1.12 [1.00; 1.59]
## 
## Test of heterogeneity:
##      Q d.f. p-value
##  11.25    9  0.2587
## 
## Results for subgroups (random effects model):
##                                                              k     SMD            95%-CI  tau^2    tau
## country:intervention = Brazil:Gender-stereotype color, ...   8  0.1091 [-0.2395; 0.4577] 0.0434 0.2082
## country:intervention = China:Gender-stereotype color,  ...   1 -0.1893 [-0.8121; 0.4335]     --     --
## country:intervention = Brazil:Gender-stereotyped motiv ...   1  0.1888 [-0.5722; 0.9498]     --     --
##                                                                Q   I^2
## country:intervention = Brazil:Gender-stereotype color, ... 10.40 32.7%
## country:intervention = China:Gender-stereotype color,  ...  0.00    --
## country:intervention = Brazil:Gender-stereotyped motiv ...  0.00    --
## 
## Test for subgroup differences (random effects model):
##                     Q d.f. p-value
## Between groups   0.83    2  0.6604
## 
## Details on meta-analytical method:
## - Inverse variance method
## - Restricted maximum-likelihood estimator for tau^2
## - Q-profile method for confidence interval of tau^2 and tau
## - Hartung-Knapp adjustment for random effects model
## - Hedges' g (bias corrected standardised mean difference; using exact formulae)
forest(m.sg4sub, digits=2, digits.sd = 2, test.overall = T, label.e = paste0(gender,':',env))

Subgroup analysis by “age:intervention”

m.sg4sub <- update.meta(m.cont, subgroup = `age:intervention`, random = T, fixed = F)
summary(m.sg4sub)
## Review:     Performance for women in stFemale
## 
##                                    SMD            95%-CI %W(random)
## S1                              0.0337 [-0.5959; 0.6633]       10.9
## S2                             -0.0214 [-0.5446; 0.5018]       14.8
## S3                             -0.5787 [-1.3128; 0.1555]        8.3
## S4                              0.6745 [-0.1048; 1.4539]        7.4
## S5                              0.2265 [-0.3774; 0.8305]       11.7
## S6                              0.4025 [-0.3214; 1.1264]        8.5
## S7                              0.5444 [-0.0754; 1.1641]       11.2
## S8: Conducted by BNU           -0.1893 [-0.8121; 0.4335]       11.1
## S9: Albuquerque, et al. (2017) -0.4198 [-1.1460; 0.3064]        8.5
## S10: Only use prompt msgs       0.1888 [-0.5722; 0.9498]        7.8
##                                                                                   age:intervention
## S1                                 adolescent:Gender-stereotype color, ranking, badges, and avatar
## S2                                 adolescent:Gender-stereotype color, ranking, badges, and avatar
## S3                                 adolescent:Gender-stereotype color, ranking, badges, and avatar
## S4                                      adult:Gender-stereotype color, ranking, badges, and avatar
## S5                                      adult:Gender-stereotype color, ranking, badges, and avatar
## S6                                      adult:Gender-stereotype color, ranking, badges, and avatar
## S7                                      adult:Gender-stereotype color, ranking, badges, and avatar
## S8: Conducted by BNU                  unknown:Gender-stereotype color, ranking, badges, and avatar
## S9: Albuquerque, et al. (2017) no-restriction:Gender-stereotype color, ranking, badges, and avatar
## S10: Only use prompt msgs               adolescent:Gender-stereotyped motivational message prompts
## 
## Number of studies combined: k = 10
## Number of observations: o = 366
## 
##                         SMD            95%-CI    t p-value
## Random effects model 0.0823 [-0.1890; 0.3537] 0.69  0.5098
## 
## Quantifying heterogeneity:
##  tau^2 = 0.0169 [0.0000; 0.4121]; tau = 0.1299 [0.0000; 0.6420]
##  I^2 = 20.0% [0.0%; 60.5%]; H = 1.12 [1.00; 1.59]
## 
## Test of heterogeneity:
##      Q d.f. p-value
##  11.25    9  0.2587
## 
## Results for subgroups (random effects model):
##                                                          k     SMD            95%-CI   tau^2    tau    Q
## age:intervention = adolescent:Gender-stereotype co ...   3 -0.1328 [-0.8798; 0.6141] <0.0001 0.0007 1.86
## age:intervention = adult:Gender-stereotype color,  ...   4  0.4401 [ 0.1343; 0.7459]       0      0 0.95
## age:intervention = unknown:Gender-stereotype color ...   1 -0.1893 [-0.8121; 0.4335]      --     -- 0.00
## age:intervention = no-restriction:Gender-stereotyp ...   1 -0.4198 [-1.1460; 0.3064]      --     -- 0.00
## age:intervention = adolescent:Gender-stereotyped m ...   1  0.1888 [-0.5722; 0.9498]      --     -- 0.00
##                                                         I^2
## age:intervention = adolescent:Gender-stereotype co ... 0.0%
## age:intervention = adult:Gender-stereotype color,  ... 0.0%
## age:intervention = unknown:Gender-stereotype color ...   --
## age:intervention = no-restriction:Gender-stereotyp ...   --
## age:intervention = adolescent:Gender-stereotyped m ...   --
## 
## Test for subgroup differences (random effects model):
##                      Q d.f. p-value
## Between groups   13.96    4  0.0074
## 
## Details on meta-analytical method:
## - Inverse variance method
## - Restricted maximum-likelihood estimator for tau^2
## - Q-profile method for confidence interval of tau^2 and tau
## - Hartung-Knapp adjustment for random effects model
## - Hedges' g (bias corrected standardised mean difference; using exact formulae)
forest(m.sg4sub, digits=2, digits.sd = 2, test.overall = T, label.e = paste0(gender,':',env))

Subgroup analysis by “ed.level:intervention”

m.sg4sub <- update.meta(m.cont, subgroup = `ed.level:intervention`, random = T, fixed = F)
summary(m.sg4sub)
## Review:     Performance for women in stFemale
## 
##                                    SMD            95%-CI %W(random)
## S1                              0.0337 [-0.5959; 0.6633]       10.9
## S2                             -0.0214 [-0.5446; 0.5018]       14.8
## S3                             -0.5787 [-1.3128; 0.1555]        8.3
## S4                              0.6745 [-0.1048; 1.4539]        7.4
## S5                              0.2265 [-0.3774; 0.8305]       11.7
## S6                              0.4025 [-0.3214; 1.1264]        8.5
## S7                              0.5444 [-0.0754; 1.1641]       11.2
## S8: Conducted by BNU           -0.1893 [-0.8121; 0.4335]       11.1
## S9: Albuquerque, et al. (2017) -0.4198 [-1.1460; 0.3064]        8.5
## S10: Only use prompt msgs       0.1888 [-0.5722; 0.9498]        7.8
##                                                                                ed.level:intervention
## S1                              upper-secundary:Gender-stereotype color, ranking, badges, and avatar
## S2                              upper-secundary:Gender-stereotype color, ranking, badges, and avatar
## S3                              upper-secundary:Gender-stereotype color, ranking, badges, and avatar
## S4                             higher-education:Gender-stereotype color, ranking, badges, and avatar
## S5                             higher-education:Gender-stereotype color, ranking, badges, and avatar
## S6                             higher-education:Gender-stereotype color, ranking, badges, and avatar
## S7                                      unknown:Gender-stereotype color, ranking, badges, and avatar
## S8: Conducted by BNU                    unknown:Gender-stereotype color, ranking, badges, and avatar
## S9: Albuquerque, et al. (2017)          unknown:Gender-stereotype color, ranking, badges, and avatar
## S10: Only use prompt msgs            upper-secundary:Gender-stereotyped motivational message prompts
## 
## Number of studies combined: k = 10
## Number of observations: o = 366
## 
##                         SMD            95%-CI    t p-value
## Random effects model 0.0823 [-0.1890; 0.3537] 0.69  0.5098
## 
## Quantifying heterogeneity:
##  tau^2 = 0.0169 [0.0000; 0.4121]; tau = 0.1299 [0.0000; 0.6420]
##  I^2 = 20.0% [0.0%; 60.5%]; H = 1.12 [1.00; 1.59]
## 
## Test of heterogeneity:
##      Q d.f. p-value
##  11.25    9  0.2587
## 
## Results for subgroups (random effects model):
##                                                               k     SMD            95%-CI   tau^2    tau
## ed.level:intervention = upper-secundary:Gender-stereoty ...   3 -0.1328 [-0.8798; 0.6141] <0.0001 0.0007
## ed.level:intervention = higher-education:Gender-stereot ...   3  0.3970 [-0.1540; 0.9481]       0      0
## ed.level:intervention = unknown:Gender-stereotype color ...   3 -0.0029 [-1.2549; 1.2491]  0.1440 0.3795
## ed.level:intervention = upper-secundary:Gender-stereoty ...   1  0.1888 [-0.5722; 0.9498]      --     --
##                                                                Q   I^2
## ed.level:intervention = upper-secundary:Gender-stereoty ... 1.86  0.0%
## ed.level:intervention = higher-education:Gender-stereot ... 0.79  0.0%
## ed.level:intervention = unknown:Gender-stereotype color ... 4.59 56.5%
## ed.level:intervention = upper-secundary:Gender-stereoty ... 0.00    --
## 
## Test for subgroup differences (random effects model):
##                     Q d.f. p-value
## Between groups   6.51    3  0.0893
## 
## Details on meta-analytical method:
## - Inverse variance method
## - Restricted maximum-likelihood estimator for tau^2
## - Q-profile method for confidence interval of tau^2 and tau
## - Hartung-Knapp adjustment for random effects model
## - Hedges' g (bias corrected standardised mean difference; using exact formulae)
forest(m.sg4sub, digits=2, digits.sd = 2, test.overall = T, label.e = paste0(gender,':',env))

Subgroup analysis by “country:age:intervention”

m.sg4sub <- update.meta(m.cont, subgroup = `country:age:intervention`, random = T, fixed = F)
summary(m.sg4sub)
## Review:     Performance for women in stFemale
## 
##                                    SMD            95%-CI %W(random)
## S1                              0.0337 [-0.5959; 0.6633]       10.9
## S2                             -0.0214 [-0.5446; 0.5018]       14.8
## S3                             -0.5787 [-1.3128; 0.1555]        8.3
## S4                              0.6745 [-0.1048; 1.4539]        7.4
## S5                              0.2265 [-0.3774; 0.8305]       11.7
## S6                              0.4025 [-0.3214; 1.1264]        8.5
## S7                              0.5444 [-0.0754; 1.1641]       11.2
## S8: Conducted by BNU           -0.1893 [-0.8121; 0.4335]       11.1
## S9: Albuquerque, et al. (2017) -0.4198 [-1.1460; 0.3064]        8.5
## S10: Only use prompt msgs       0.1888 [-0.5722; 0.9498]        7.8
##                                                                                  country:age:intervention
## S1                                 Brazil:adolescent:Gender-stereotype color, ranking, badges, and avatar
## S2                                 Brazil:adolescent:Gender-stereotype color, ranking, badges, and avatar
## S3                                 Brazil:adolescent:Gender-stereotype color, ranking, badges, and avatar
## S4                                      Brazil:adult:Gender-stereotype color, ranking, badges, and avatar
## S5                                      Brazil:adult:Gender-stereotype color, ranking, badges, and avatar
## S6                                      Brazil:adult:Gender-stereotype color, ranking, badges, and avatar
## S7                                      Brazil:adult:Gender-stereotype color, ranking, badges, and avatar
## S8: Conducted by BNU                   China:unknown:Gender-stereotype color, ranking, badges, and avatar
## S9: Albuquerque, et al. (2017) Brazil:no-restriction:Gender-stereotype color, ranking, badges, and avatar
## S10: Only use prompt msgs               Brazil:adolescent:Gender-stereotyped motivational message prompts
## 
## Number of studies combined: k = 10
## Number of observations: o = 366
## 
##                         SMD            95%-CI    t p-value
## Random effects model 0.0823 [-0.1890; 0.3537] 0.69  0.5098
## 
## Quantifying heterogeneity:
##  tau^2 = 0.0169 [0.0000; 0.4121]; tau = 0.1299 [0.0000; 0.6420]
##  I^2 = 20.0% [0.0%; 60.5%]; H = 1.12 [1.00; 1.59]
## 
## Test of heterogeneity:
##      Q d.f. p-value
##  11.25    9  0.2587
## 
## Results for subgroups (random effects model):
##                                                                  k     SMD            95%-CI   tau^2    tau
## country:age:intervention = Brazil:adolescent:Gender-stereo ...   3 -0.1328 [-0.8798; 0.6141] <0.0001 0.0007
## country:age:intervention = Brazil:adult:Gender-stereotype  ...   4  0.4401 [ 0.1343; 0.7459]       0      0
## country:age:intervention = China:unknown:Gender-stereotype ...   1 -0.1893 [-0.8121; 0.4335]      --     --
## country:age:intervention = Brazil:no-restriction:Gender-st ...   1 -0.4198 [-1.1460; 0.3064]      --     --
## country:age:intervention = Brazil:adolescent:Gender-stereo ...   1  0.1888 [-0.5722; 0.9498]      --     --
##                                                                   Q  I^2
## country:age:intervention = Brazil:adolescent:Gender-stereo ... 1.86 0.0%
## country:age:intervention = Brazil:adult:Gender-stereotype  ... 0.95 0.0%
## country:age:intervention = China:unknown:Gender-stereotype ... 0.00   --
## country:age:intervention = Brazil:no-restriction:Gender-st ... 0.00   --
## country:age:intervention = Brazil:adolescent:Gender-stereo ... 0.00   --
## 
## Test for subgroup differences (random effects model):
##                      Q d.f. p-value
## Between groups   13.96    4  0.0074
## 
## Details on meta-analytical method:
## - Inverse variance method
## - Restricted maximum-likelihood estimator for tau^2
## - Q-profile method for confidence interval of tau^2 and tau
## - Hartung-Knapp adjustment for random effects model
## - Hedges' g (bias corrected standardised mean difference; using exact formulae)
forest(m.sg4sub, digits=2, digits.sd = 2, test.overall = T, label.e = paste0(gender,':',env))

Subgroup analysis by “country:ed.level:intervention”

m.sg4sub <- update.meta(m.cont, subgroup = `country:ed.level:intervention`, random = T, fixed = F)
summary(m.sg4sub)
## Review:     Performance for women in stFemale
## 
##                                    SMD            95%-CI %W(random)
## S1                              0.0337 [-0.5959; 0.6633]       10.9
## S2                             -0.0214 [-0.5446; 0.5018]       14.8
## S3                             -0.5787 [-1.3128; 0.1555]        8.3
## S4                              0.6745 [-0.1048; 1.4539]        7.4
## S5                              0.2265 [-0.3774; 0.8305]       11.7
## S6                              0.4025 [-0.3214; 1.1264]        8.5
## S7                              0.5444 [-0.0754; 1.1641]       11.2
## S8: Conducted by BNU           -0.1893 [-0.8121; 0.4335]       11.1
## S9: Albuquerque, et al. (2017) -0.4198 [-1.1460; 0.3064]        8.5
## S10: Only use prompt msgs       0.1888 [-0.5722; 0.9498]        7.8
##                                                                               country:ed.level:intervention
## S1                              Brazil:upper-secundary:Gender-stereotype color, ranking, badges, and avatar
## S2                              Brazil:upper-secundary:Gender-stereotype color, ranking, badges, and avatar
## S3                              Brazil:upper-secundary:Gender-stereotype color, ranking, badges, and avatar
## S4                             Brazil:higher-education:Gender-stereotype color, ranking, badges, and avatar
## S5                             Brazil:higher-education:Gender-stereotype color, ranking, badges, and avatar
## S6                             Brazil:higher-education:Gender-stereotype color, ranking, badges, and avatar
## S7                                      Brazil:unknown:Gender-stereotype color, ranking, badges, and avatar
## S8: Conducted by BNU                     China:unknown:Gender-stereotype color, ranking, badges, and avatar
## S9: Albuquerque, et al. (2017)          Brazil:unknown:Gender-stereotype color, ranking, badges, and avatar
## S10: Only use prompt msgs            Brazil:upper-secundary:Gender-stereotyped motivational message prompts
## 
## Number of studies combined: k = 10
## Number of observations: o = 366
## 
##                         SMD            95%-CI    t p-value
## Random effects model 0.0823 [-0.1890; 0.3537] 0.69  0.5098
## 
## Quantifying heterogeneity:
##  tau^2 = 0.0169 [0.0000; 0.4121]; tau = 0.1299 [0.0000; 0.6420]
##  I^2 = 20.0% [0.0%; 60.5%]; H = 1.12 [1.00; 1.59]
## 
## Test of heterogeneity:
##      Q d.f. p-value
##  11.25    9  0.2587
## 
## Results for subgroups (random effects model):
##                                                                       k     SMD            95%-CI   tau^2
## country:ed.level:intervention = Brazil:upper-secundary:Gender-s ...   3 -0.1328 [-0.8798; 0.6141] <0.0001
## country:ed.level:intervention = Brazil:higher-education:Gender- ...   3  0.3970 [-0.1540; 0.9481]       0
## country:ed.level:intervention = Brazil:unknown:Gender-stereotyp ...   2  0.0816 [-6.0390; 6.2023]  0.3462
## country:ed.level:intervention = China:unknown:Gender-stereotype ...   1 -0.1893 [-0.8121; 0.4335]      --
## country:ed.level:intervention = Brazil:upper-secundary:Gender-s ...   1  0.1888 [-0.5722; 0.9498]      --
##                                                                        tau    Q   I^2
## country:ed.level:intervention = Brazil:upper-secundary:Gender-s ... 0.0007 1.86  0.0%
## country:ed.level:intervention = Brazil:higher-education:Gender- ...      0 0.79  0.0%
## country:ed.level:intervention = Brazil:unknown:Gender-stereotyp ... 0.5884 3.92 74.5%
## country:ed.level:intervention = China:unknown:Gender-stereotype ...     -- 0.00    --
## country:ed.level:intervention = Brazil:upper-secundary:Gender-s ...     -- 0.00    --
## 
## Test for subgroup differences (random effects model):
##                     Q d.f. p-value
## Between groups   7.50    4  0.1117
## 
## Details on meta-analytical method:
## - Inverse variance method
## - Restricted maximum-likelihood estimator for tau^2
## - Q-profile method for confidence interval of tau^2 and tau
## - Hartung-Knapp adjustment for random effects model
## - Hedges' g (bias corrected standardised mean difference; using exact formulae)
forest(m.sg4sub, digits=2, digits.sd = 2, test.overall = T, label.e = paste0(gender,':',env))

Funnel Plot

m.cont <- update.meta(m.cont, studlab = data$study)
summary(eggers.test(x = m.cont))
## Eggers' test of the intercept 
## ============================= 
## 
##  intercept       95% CI     t   p
##      0.397 -5.54 - 6.33 0.131 0.9
## 
## Eggers' test does not indicate the presence of funnel plot asymmetry.
funnel(m.cont, xlab = "Hedges' g", studlab = T, legend=T, addtau2 = T)